CN112214725A

CN112214725A - Serial FFT (fast Fourier transform) co-processing method, digital oscilloscope and storage medium

Info

Publication number: CN112214725A
Application number: CN202011419854.5A
Authority: CN
Inventors: 陈报; 郑文明; 周旭鑫
Original assignee: Shenzhen Siglent Technologies Co Ltd
Current assignee: Shenzhen Siglent Technologies Co Ltd
Priority date: 2020-12-08
Filing date: 2020-12-08
Publication date: 2021-01-12
Anticipated expiration: 2040-12-08
Also published as: CN112214725B

Abstract

The application relates to a serial FFT coprocessing method, a digital oscilloscope and a storage medium, wherein the serial FFT coprocessing method comprises the following steps: carrying out conversion processing on the sampled data of the signal by using an FFT (fast Fourier transform) accelerated computing kernel to obtain conversion data; butterfly index information and phase factor information of the sampling data are generated; and searching transformation data by using butterfly index information, and performing butterfly operation on the index result and the phase factor information to obtain an operation result. Because the butterfly index information and the phase factor information of the sampled data are generated, the data are easily traversed and transformed in an index mode, thereby realizing butterfly operation with the phase factor and further improving the speed of performing time domain-frequency domain transformation on the sampled data; the technical scheme can support FFT operation of a large number of data points, only needs to reasonably configure the length of the data points and calculate the grading sequence, and flexible configuration means are easy to adapt to different storage depths of the oscilloscope.

Description

Serial FFT (fast Fourier transform) co-processing method, digital oscilloscope and storage medium

Technical Field

The invention relates to the technical field of oscilloscopes, in particular to a serial FFT coprocessing method, a digital oscilloscope and a storage medium.

Background

Oscilloscopes are indispensable tools for designing, manufacturing and maintaining electronic equipment, most of the existing oscilloscopes mainly use digital oscilloscopes, are increasingly popularized due to functions of waveform triggering, storing, displaying, measuring, analyzing and the like, and with rapid development of scientific and market requirements, the digital oscilloscopes are considered to be eyes of engineers and are used as necessary tools for meeting measurement challenges of the engineers.

In the current oscilloscope, Fast Fourier Transform (FFT) is a very powerful analysis function, and based on advanced FFT analysis, designers can accurately know interference signal frequency points, signal power spectrum, signal frequency composition, and the frequency cut-off characteristics of a filter circuit, etc. introduced into signals. However, limited by the computing power of the oscilloscope, the FFT function of the oscilloscope does not exert due effects, and the FFT analysis application of the oscilloscope is seriously affected by the disadvantages of slow computing speed, long dead time and the like in large-point analysis.

Disclosure of Invention

The invention mainly solves the technical problems that: how to improve the FFT analysis capability of the oscilloscope. In order to solve the above problems, the present application provides a serial FFT co-processing method, a digital oscilloscope, and a storage medium.

According to a first aspect, an embodiment provides a serial FFT co-processing method, which includes: carrying out conversion processing on the sampled data of the signal by using an FFT (fast Fourier transform) accelerated computing kernel to obtain conversion data; butterfly index information and phase factor information of the sampling data are generated; searching the transformation data by using the butterfly index information, and performing butterfly operation on an index result and the phase factor information to obtain an operation result; the operation result is used for frequency domain analysis of the signal.

The transforming processing of the sample data of the signal by using the FFT acceleration calculation kernel to obtain the transformed data comprises the following steps: determining a calculation grading order according to the number of data points of the sampling data and the number of input points of the FFT acceleration calculation kernel; reordering each data point in the sampled data according to an input data hierarchical ordering rule of the FFT acceleration calculation kernel to obtain data index information corresponding to the highest hierarchy in the calculated hierarchical order; and searching the sampling data by using the data index information, grouping index results, performing fast Fourier transform according to the groups, and arranging the transform results according to a natural sequence to obtain transform data.

The generating butterfly index information and phase factor information of the sample data includes: reordering the sampling sequence of the sampling data according to a preset butterfly operation index rule to obtain a butterfly index corresponding to each grading, and integrating butterfly index information of the sampling data; and taking the sequence of each grading as an independent variable, generating a phase factor corresponding to each grading according to a preset phase factor calculation rule, and integrating to form phase factor information of the sampling data.

The butterfly operation index rule comprises: setting positive integer exponential power as step length, combining every two even number serial numbers or every two odd number serial numbers in a sampling sequence of the sampling data, wherein the difference value of the serial numbers in each combination is the step length, and arranging and distributing the formed combinations according to a natural sequence; the phase factor calculation rule includes: setting the sequence of each grading as k, inputting the formulaPhase_factor= exp(-i×2π/2^m×(0:(2^(m-k) -1)) ×2^k) So as to calculate and obtain a corresponding phase factor; wherein the content of the first and second substances,exp() In order to operate in an exponential manner,iin units of imaginary numbers, (0 (2(m-k) -1)) is an increasing vector starting from 0 to 2(m-k) -1, and m is a power number of the sampled data when the number of data points is base on 2.

The searching the transformation data by using the butterfly index information, and performing butterfly operation on the index result and the phase factor information to obtain an operation result, including: according to the calculated grading order, searching the transformation data by using the butterfly index corresponding to each grading to obtain a plurality of groups of index data, and searching the phase factor information by using the butterfly index corresponding to each grading to obtain phase factors respectively matched with each group of index data; multiplying each group of index data by the matched phase factor through butterfly operation to obtain product data corresponding to each grading; and sequentially carrying out butterfly operation from the highest grade in the grading order to obtain product data corresponding to each grade until iteration is carried out to the lowest grade in the grading order, and taking the obtained product data as an operation result.

According to a second aspect, an embodiment provides a digital oscilloscope comprising a first processing component, a second processing component, and a memory communicatively coupled to each other; the first processing component is used for acquiring sampling data of a signal, performing conversion processing on the sampling data by utilizing an FFT (fast Fourier transform) accelerated computing kernel to generate butterfly index information of the sampling data, and storing the conversion data obtained by the conversion processing and the butterfly index information of the sampling data into the memory; the second processing component is used for generating phase factor information of the sampling data, searching the transformation data by using the butterfly index information in the memory, performing butterfly operation on an index result and the phase factor information, processing to obtain an operation result, and storing the operation result into the memory; the first processing unit is further configured to obtain the operation result from the memory, and perform frequency domain analysis on the operation result to form an analysis result.

The first processing component comprises a calculation grading module, a first data sorting module, a second data sorting module and a transformation processing module; the calculation grading module is used for determining a calculation grading order according to the number of data points of the sampling data and the number of input points of the FFT acceleration calculation kernel; the first data sorting module is used for re-sorting each data point in the sampled data according to the input data hierarchical sorting rule of the FFT accelerated computing kernel to obtain data index information corresponding to the highest hierarchical level in the computed hierarchical order; the transformation processing module is used for searching the sampling data by using the data index information, grouping index results and performing fast Fourier transformation according to groups to obtain a transformation result corresponding to each group; and the second data sorting module is used for re-sorting the sampling sequence of the sampling data according to a preset butterfly operation index rule to obtain a butterfly index corresponding to each grade.

A first storage space and a second storage space are arranged in the memory; the first storage space comprises a plurality of designated areas corresponding to the hierarchical order, the designated areas are used for sequentially storing the transformation results corresponding to the groups generated by the transformation processing module, and the transformation data are formed by arranging the natural order of the data stored in the designated areas; the second storage space comprises a plurality of index regions corresponding to the hierarchical order; the index areas are used for sequentially storing each butterfly index corresponding to the grades generated by the second data sorting module, and the butterfly index information is formed by the natural sequence arrangement of the written data in each index area.

The digital oscilloscope further comprises a display; the display is connected with the first processing component and used for displaying the analysis result formed by the first processing component.

According to a third aspect, the invention provides a computer readable storage medium comprising a program executable by a processor to implement the method described in the first aspect above.

The invention has the beneficial effects that:

the serial FFT co-processing method, the digital oscilloscope and the storage medium provided by the above embodiments, wherein the serial FFT co-processing method includes: carrying out conversion processing on the sampled data of the signal by using an FFT (fast Fourier transform) accelerated computing kernel to obtain conversion data; butterfly index information and phase factor information of the sampling data are generated; and searching transformation data by using butterfly index information, and performing butterfly operation on the index result and the phase factor information to obtain an operation result. On the first hand, the idea of sampling serial processing in the technical scheme is that the serial calculation is performed on the sampled data by selecting an FFT (fast Fourier transform) accelerated calculation inner core, so that the condition that huge multiplication and addition resources are needed in the prior parallel processing can be avoided, and the calculation performance of equipment is improved; in the second aspect, because the butterfly index information and the phase factor information of the sampled data are generated, the data are easily traversed and transformed in an index mode, thereby realizing butterfly operation with the phase factor and further improving the speed of time domain-frequency domain transformation of the sampled data; in the third aspect, the technical scheme can support FFT operation of a large number of data points, only the length of the data points and the calculation grading sequence need to be reasonably configured, and the flexible configuration means is easy to adapt to different storage depths of the oscilloscope; in the fourth aspect, when the serial FFT co-processing method is applied to the digital oscilloscope, a first processing component, a second processing component and a memory which are mutually communicated and connected are arranged to execute different processing processes, the first processing component is used for finishing the work of sequencing and data sending, and the second processor is used for finishing the work of phase factor generation and butterfly operation, so that the processing pressure of a single logic device is favorably differentiated, and the overall operation efficiency of the equipment is improved; in the fifth aspect, the technical scheme reasonably distributes the serial FFT coprocessing tasks by using different functional devices, and arranges the complex data sequencing work in the first processor for execution, thereby easily realizing the design idea of avoiding heavy matters and reducing light weight, and fully exerting the data serial FFT processing capability of the digital oscilloscope.

Drawings

FIG. 1 is a schematic structural diagram of a digital oscilloscope according to an embodiment of the present application;

FIG. 2 is a schematic structural view of a first processing element;

FIG. 3 is a schematic diagram of a second processing element;

fig. 4 is a flowchart of a serial FFT co-processing method according to a second embodiment of the present application;

FIG. 5 is a flow chart of the transformation process resulting in transformed data;

FIG. 6 is a flow chart for generating butterfly index information and phase factor information;

FIG. 7 is a flow chart of a butterfly operation performed on the index result and the phase factor information;

FIG. 8 is a schematic diagram of a data ordering rule in one embodiment;

FIG. 9 is a schematic diagram of an embodiment of a butterfly index rule;

FIG. 10 is a schematic diagram of a spy shape operation in one embodiment;

FIG. 11 is a schematic diagram of memory space partitioning in one embodiment;

FIG. 12 is a schematic diagram of the writing and reading of memory data in one embodiment;

fig. 13 is a schematic structural diagram of a digital oscilloscope in the third embodiment of the present application.

Detailed Description

The present invention will be described in further detail with reference to the following detailed description and accompanying drawings. Wherein like elements in different embodiments are numbered with like associated elements. In the following description, numerous details are set forth in order to provide a better understanding of the present application. However, those skilled in the art will readily recognize that some of the features may be omitted or replaced with other elements, materials, methods in different instances. In some instances, certain operations related to the present application have not been shown or described in detail in order to avoid obscuring the core of the present application from excessive description, and it is not necessary for those skilled in the art to describe these operations in detail, so that they may be fully understood from the description in the specification and the general knowledge in the art.

Furthermore, the features, operations, or characteristics described in the specification may be combined in any suitable manner to form various embodiments. Also, the various steps or actions in the method descriptions may be transposed or transposed in order, as will be apparent to one of ordinary skill in the art. Thus, the various sequences in the specification and drawings are for the purpose of describing certain embodiments only and are not intended to imply a required sequence unless otherwise indicated where such sequence must be followed.

The numbering of the components as such, e.g., "first", "second", etc., is used herein only to distinguish the objects as described, and does not have any sequential or technical meaning. The term "connected" and "coupled" when used in this application, unless otherwise indicated, includes both direct and indirect connections (couplings).

For clear and accurate understanding of the technical solutions of the present application, some technical terms will be described herein.

Fft (fast Fourier transform), is a fast algorithm for Discrete Fourier Transform (DFT). The fast Fourier transform is obtained by improving the algorithm of the discrete Fourier transform according to the characteristics of odd, even, virtual, real and the like of the discrete Fourier transform, and the corresponding frequency domain signal can be obtained by carrying out FFT processing on the time domain signal in the invention.

The technical solution of the present application will be specifically described with reference to the following examples.

The first embodiment,

Referring to fig. 1 to 3, the present embodiment discloses a digital oscilloscope, which includes a first processing unit 11, a second processing unit 12 and a memory 13, which are communicatively connected to each other, and are described below.

The first processing unit 11 may adopt a processor (CPU) and is mainly configured to obtain sample data of a signal, perform transform processing on the sample data by using an FFT accelerated computation kernel, generate butterfly index information of the sample data, and store the transform data obtained by the transform processing and the butterfly index information of the sample data in the memory 13. The FFT acceleration calculation kernel can adopt an Xilinx FFT ip kernel, for example, input points of 1024 points are executed, and then FFT results of 1024-4M points are calculated according to a set algorithm.

The second processing unit 12 may adopt a programmable logic processing unit (FPGA), and is mainly configured to generate phase factor information of the sampled data, search the transformed data by using butterfly index information in the memory 13, perform butterfly operation on the index result and the phase factor information, process the index result and the phase factor information to obtain an operation result, and store the operation result in the memory 13.

In addition, the first processing unit 11 is further configured to obtain the operation result from the memory 13, and perform frequency domain analysis on the operation result to form an analysis result. It should be noted that frequency domain analysis is a classical signal analysis method, and is to apply an illustration to evaluate signal performance in a frequency domain, and the specifically adopted method is also various, such as analysis of frequency spectrum, energy spectrum, power spectrum, cepstrum, wavelet, and the like, and even information (possibly amplitude, power, intensity, phase, and the like) of a signal under different frequencies can be found.

In one embodiment, referring to fig. 2, the first processing component 11 includes a computational ranking module 111, a first data ordering module 112, a second data ordering module 114, and a transformation processing module 113, each described below.

The calculation ranking module 111 is configured to determine a ranking order of the calculation according to the number of data points of the sample data and the number of input points of the FFT acceleration calculation kernel. For example, the number of data points is 4M =2²²If the number of input points of the FFT-accelerated computation kernel is 1024=2¹⁰Then there are 12 remaining stages (i.e. 12 stages) to be calculated and the ranking order is reduced from 12 to 1.

The first data sorting module 112 is configured to reorder, according to the input data hierarchical sorting rule of the FFT accelerated computation kernel, each data point in the sampled data to obtain data index information corresponding to the highest hierarchy. For example, for the sample data, it can be expressed as x (n) = [ x (0), x (1), … … x (n-1), x (n)]And n is equal to [0, 2 ]²²-1](ii) a If a function search _ index _ function (data, stage) is used to indicate the input data ranking rule of the FFT computation-accelerated kernel, the data _ index computed when data = x (n) and stage =12 indicates the data index information corresponding to the highest ranking.

The transform processing module 113 is configured to search for sample data by using the data index information, group the index results, and perform fast fourier transform on the index results according to the groups to obtain a transform result corresponding to each group. It should be noted that the data index information represents an index result obtained by alternately arranging the sampling order of each data point in the sampled data according to the odd number and the even number, so that the data of the odd number column or the even number column is easily found through the data index information, and for the FFT accelerated computation core with the input point number of 1024 points, 1024 data corresponding to each odd number column or even number column can be divided into one group, so as to perform fast fourier transform (FFT processing) on the group.

The second data sorting module 114 is configured to reorder the sampling sequence of the sampling data according to a preset butterfly operation index rule to obtain each hierarchyThe corresponding butterfly index. The butterfly index rule herein includes: setting positive integer exponentiation (e.g. 2)ⁿN is a positive integer) as a step length, combining every two even-numbered serial numbers or every two odd-numbered serial numbers in a sampling sequence of the sampling data, and taking a serial number difference value in each combination as the step length, and arranging and distributing the formed combinations according to a natural sequence. After the butterfly index corresponding to each grade is obtained, the butterfly index information of the sampling data is easily integrated.

In one embodiment, referring to fig. 3, the second processing unit 12 includes a data input buffer module 121, an FFT accelerated computation core 122, an address controller 124, a butterfly controller 123, a phase factor generator 125, an FFT output buffer module 126, and an FFT input buffer module 127.

In order to reduce the processing load of the first processing unit 11 (e.g., CPU), the FFT acceleration computation core may be disposed in the first processing unit 12 (e.g., FPGA); then, the data input buffer module 121 is configured to buffer the sampling data sent by the first processing unit 11, and only do buffering and do not perform operation; therefore, the FFT accelerated computation kernel 122 transforms the sample data buffered by the data input buffer module 121, obtains transformed data, and stores the transformed data in the memory 13. The user may select 2 based on system speed and resource requirementsⁿFFT of size to accelerate the computation core 122, preferably using 1024=2¹⁰And inputting a calculation kernel of the point number.

Wherein the phase factor generator 125 is used to generate phase factor information of the sampled data. For example, the order of each classification is used as an independent variable, a phase factor corresponding to each classification is generated through a preset phase factor calculation rule, and phase factor information of the sampling data is integrated. The phase factor calculation rule includes: setting the sequence of each grading as k, inputting the formulaPhase_factor= exp(-i×2π/2^m×(0:(2^(m-k) -1)) ×2^k) So as to calculate and obtain a corresponding phase factor; wherein the content of the first and second substances,exp() In order to operate in an exponential manner,iis an imaginary unit, (0 (2)^(m-k)-1)) from 0 to 2^(m-k)-1 increment vector, m being the sampled dataThe number of data points (2) is a power number of the base; if the number of data points is 4M =2²²Then m = 22.

The butterfly operation controller 123 is configured to search the transform data in the memory 13 by using the butterfly index information in the memory 13, and perform a butterfly operation on the index result and the phase factor information generated by the phase factor generator 125 to obtain an operation result and store the operation result in the memory 13.

The address controller 124 is connected to the FFT acceleration computation core 122, the butterfly controller 123 and the controller 13, and is configured to perform address control on access data of the memory 13, so as to facilitate data storage and data retrieval operations among the FFT computation core 122, the butterfly controller 123 and the memory 13. The FFT output buffer module 126 is used for buffering the operation result generated by the butterfly operation controller 123 through the butterfly operation, and controlling the operation result to be stored in the memory 13. The FFT input buffer module 127 is configured to read the butterfly index information and the transform data from the memory 13, and transmit the butterfly index information and the transform data to the butterfly operation controller 123 after buffering.

In one embodiment, a first storage space and a second storage space are provided in the memory 13. Then, the first storage space includes a plurality of designated areas corresponding to the hierarchical order, and the plurality of designated areas are used for sequentially storing the transformation results corresponding to the respective groups generated by the transformation processing module 113 in fig. 2, and the transformation data is configured by arranging the data stored in the respective designated areas in a natural order. The second storage space includes a plurality of index regions corresponding to the hierarchical order, and the plurality of index regions are used to sequentially store each hierarchical butterfly index generated by the second data sorting module 114 in fig. 2, and butterfly index information is formed by natural sequential arrangement of data written in each index region.

In one embodiment, the digital oscilloscope of FIG. 1 further comprises a display (not illustrated in FIG. 1). The display is connected to the first processing unit 11 and is used for displaying the analysis result formed by the first processing unit 11, such as displaying information of a spectrogram, a power spectrum, and the like of the signal, which is not limited herein.

Those skilled in the art can understand that, in this embodiment, when the serial FFT co-processing method is applied to the digital oscilloscope, the first processing unit, the second processing unit, and the memory, which are communicatively connected to each other, are configured to execute different processing procedures, the first processing unit is used to complete the operations of sorting and sending data, and the second processor is used to complete the operations of phase factor generation and butterfly operation, which is helpful to differentiate the processing pressure of a single logic device, thereby improving the overall operation efficiency of the device. In addition, the technical scheme reasonably distributes serial FFT coprocessing tasks by using different functional devices, and arranges the complex data sequencing work in the first processor for execution, thereby easily realizing the design idea of avoiding heavy matters and reducing light matters, and fully exerting the data serial FFT processing capability of the digital oscilloscope.

Example II,

On the basis of the digital oscilloscope disclosed in the first embodiment, the present embodiment discloses a serial FFT co-processing method for a digital oscilloscope, which is mainly used for the first processing unit 11 and the second processing unit 12 in the first embodiment.

Referring to fig. 4, the disclosed serial FFT co-processing method mainly includes steps S210-S230, which are described below.

Step S210, transform the sampled data of the signal by using the FFT accelerated computation kernel to obtain transformed data. For example, as shown in fig. 3, after the first processing unit 11 acquires the sample data of the signal, the sample data is sent to the second processing unit 12, so that the FFT-accelerated computation kernel in the second processing unit 12 performs the transform processing on the sample data, and the transform data obtained by the transform processing is stored in the memory 13.

Step S220, butterfly index information and phase factor information of the sample data are generated. Referring to fig. 2 and 3, the calculation classification module 111 in the first processing component 11 may be used to determine the calculation classification order according to the number of data points of the sample data and the number of input points of the FFT accelerated computation kernel, so that the second data sorting module 114 reorders the sample sequence of the sample data according to the preset butterfly operation index rule to obtain the butterfly index corresponding to each classification, and the butterfly index information is integrated and stored in the memory 13.

Further, the phase factor generator 125 in the second processing section 12 generates phase factor information of the sampled data. For example, the phase factor generator 125 uses the order of each classification as an argument, generates a phase factor corresponding to each classification according to a preset phase factor calculation rule, and integrates phase factor information forming the sampling data.

Step S230, searching transformation data by using butterfly index information, and performing butterfly operation on the index result and the phase factor information to obtain an operation result; the result of the operation is used for frequency domain analysis of the signal. For example, as shown in fig. 3, the butterfly operation controller 123 in the second processing component 12 obtains butterfly index information from the memory 13 through the FFT input buffer module 127, searches the transformed data in the memory 13, multiplies the data corresponding to the butterfly index by the phase factor information generated by the phase factor generator 125, and obtains an operation result through butterfly operation.

In this embodiment, referring to fig. 5, the step S210 mainly relates to the process of transforming the transformed data, and may specifically include steps S211 to S213, which are respectively described as follows.

In step S211, the calculation ranking module 111 of the first processing unit 11 in fig. 2 determines the ranking order of calculation according to the number of data points of the sample data and the number of input points of the FFT-accelerated calculation kernel. Because the number of input points of the FFT acceleration calculation kernel is fixed, a plurality of sampling data are subjected to fast Fourier transform in a serial form, and the sampling data of one frame are input into the FFT acceleration calculation kernel for a plurality of times, so that a plurality of grades are formed.

For example, if the sample data is represented by x (n) = [ x (0), x (1), … … x (n-1), x (n)]，n∈[0, 2²²-1]The number of data points for each frame of sampled data is then 4M =2²²The input point number of the FFT acceleration calculation kernel is 1024=2¹⁰Then the number of grades can be calculated as 12, and then the sampled data will be fast fourier transformed in 12 grading orders.

As another example, if the sampled data is represented as x (n) = [ x (0), x(1), …… x(14), x(15)]the number of data points for the sampled data per frame is then 16=2⁴The input point number of the FFT acceleration calculation kernel is 4=2²Then the computed number of levels can be 2, and then the sample data will be fast fourier transformed in 2 hierarchical orders.

In step S212, the first sorting module 111 of the first processing unit 11 in fig. 2 reorders each data point in the sampled data according to the input data hierarchical sorting rule of the FFT accelerated computation kernel, so as to obtain data index information corresponding to the highest hierarchy.

In a specific embodiment, the input data hierarchical ordering rule of the FFT-accelerated computation kernel may refer to fig. 8, and sample data x (n) = [ x (0), x (1), … … x (14), x (15) ] of 16 points is illustrated as an example. In fig. 8, p denotes an original sample sequence of one frame of sample data, and 0 to 15 denote sample sequences of sample data.

When ranked as 1 (i.e., stage = 1), the sequence p will be divided into 8-dot even columns and 8-dot odd columns, and expressed as 8 dots, respectively

x (even) = [ x (0), x (2), x (4), x (6), x (8), x (10), x (12), x (14) ];

x (odd) = [ x (1), x (3), x (5), x (7), x (9), x (11), x (13), x (15) ].

When the gradation is 2 (i.e., stage = 2), the even-numbered columns of 8 dots can be divided into the even-numbered columns of 4 dots, and the odd-numbered columns of 8 dots can be divided into the odd-numbered columns of 4 dots, and are respectively expressed as

x(n0)=[x(0),x(4),x(8),x(12)]；

x(n1)=[x(2),x(6),x(10),x(14)]；

x(n2)=[x(1),x(5),x(9),x(13)]；

x(n3)=[x(3),x(7),x(11),x(15)]。

In the case of classification 3 (i.e., stage = 3), with 2 data points as the minimum unit, it is possible to further classify the classification into one level, so as to obtain the ranking result: [ x (0), x (8) ], [ x (4), x (12) ], [ x (2), x (10) ], [ x (6), x (14) ] [ x (1), x (9) ], [ x (5), x (13) ] [ x (3), x (11) ], [ x (7), x (15) ].

Then, according to the schematic content in fig. 8, the input data ranking rule of the FFT accelerated computation kernel may be represented by a function search _ index _ function (data, stage), and only the input sample data = x (n) and the highest ranking stage are required to automatically return the data index information search _ index corresponding to the highest ranking. If x (n) = [ x (0), x (1), … … x (14), x (15) ] and highest ranked stage =2 are input to the function, the returned data index information is represented as: 0. 4, 8, 12; 2. 6, 10, 14; 1. 5, 9, 13; 3. 7, 11 and 15.

In step S213, the transform processing module 113 of the first processing unit 11 in fig. 2 searches for sample data using the data index information, groups the index results and performs fast fourier transform on the group results, and arranges the transform results in a natural order, thereby obtaining transform data.

Referring to fig. 2 and 3, for sample data x (n) = [ x (0), x (1), … … x (14), x (15) ] of 16 points, data index information (0, 4, 8, 12; 2, 6, 10, 14; 1, 5, 9, 13; 3, 7, 11, 15) corresponding to the highest-ranking stage =2 is stored in the memory 13, then the transform processing module 113 searches for the sample data using the data index information, and the search results are divided into four groups and respectively represented as four groups

x(n0)=[x(0),x(4),x(8),x(12)]，x(n1)=[x(2),x(6),x(10),x(14)]，x(n2)=[x(1),x(5),x(9),x(13)]，x(n3)=[x(3),x(7),x(11),x(15)]。

The transformation processing module 113 sends each set of results x (n0), x (n1), x (n2), and x (n3) to the second processing unit 12, respectively, obtains corresponding transformation results through fast fourier transformation, respectively, and obtains transformation data after storing the transformation results in the memory 13 and arranging the transformation results in a natural order.

In this embodiment, referring to fig. 6, the step S220 mainly relates to a process of generating butterfly index information and phase factor information of the sample data, and may specifically include steps S221 to S222, which are respectively described as follows.

In step S221, the second sorting module 114 of the first processing component 11 in fig. 2 reorders the sampling sequence of the sampling data according to the preset butterfly operation index rule, so as to obtain the butterfly index corresponding to each stage, and integrate the butterfly index information forming the sampling data.

In one implementation, the butterfly index rule includes: setting a base 2 positive integer exponential power (e.g., 2)ⁿ) And as the step length, combining every two even-numbered or every two odd-numbered in the sampling sequence of the sampling data, wherein the difference value of the numbers in each combination is the step length, and the formed combinations are arranged and distributed according to a natural sequence.

For example, the butterfly operation index rule shown in fig. 9 is to set 16-point sample data x (n) = [ x (0), x (1), … … x (14), x (15)]This is illustrated as an example. When the grade is 3 (i.e. stage = 3), the grade is 2¹As step sizes, the result of combining every two even ordinal numbers or every two odd ordinal numbers is: 0. 2,1, 3, 4, 6, 5, 7, 8, 10, 9, 11, 12, 14, 13 and 15, which are arranged and distributed according to a natural sequence to form butterfly index information. When the grade is 2 (i.e. stage = 2), the grade is 2²As step sizes, the result of combining every two even ordinal numbers or every two odd ordinal numbers is: 0. 4,1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11 and 15 are arranged and distributed according to a natural sequence to form butterfly index information. When the grade is 1 (i.e. stage = 1), the grade is 2³As step sizes, the result of combining every two even ordinal numbers or every two odd ordinal numbers is: 0. 8, 1, 9, 2, 10, 3, 11, 4, 12, 5, 13, 6, 14, 7 and 15 are arranged and distributed according to a natural sequence to form butterfly index information.

Then, according to the schematic content in fig. 9, the function button _ index _ function (DATA, stage) may be used to represent the input DATA hierarchical ordering rule of the FFT accelerated computation kernel, and only the original sample sequence DATA and the butterfly index button _ index corresponding to each hierarchical stage need to be input, and automatically returned. If DATA = [0,1,2 … … 13,14,15] and hierarchical stage =2 are input to the function, the butterfly index returned is expressed as: 0. 4,1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11, 15.

In step S222, the phase factor generator 125 of the second processing unit in fig. 3 uses the order of each classification as an argument, generates a phase factor corresponding to each classification according to a preset phase factor calculation rule, and integrates the phase factor information of the sampled data.

In one embodiment, the preset phase factor calculation rule includes: setting the sequence of each grading as k, inputting the formulaPhase_factor= exp(-i×2π/2^m×(0:(2^(m-k) -1)) ×2^k) So as to calculate and obtain a corresponding phase factor; wherein the content of the first and second substances,exp() In order to operate in an exponential manner,iin the unit of imaginary number, (0 (2(m-k) -1)) is an increasing vector starting from 0 to 2(m-k) -1, and m is a power number when the number of data points of the sample data is set to 2. For example, in fig. 10, 16-point sample data x (n) = [ x (0), x (1), … … x (14), x (15)]As an example, the number of data points of the sample data is 16=2⁴Then m = 4; then, in the case of stage =2 (i.e., k = 2) in the hierarchy, the calculated phase factors are respectively expressed as: e (-i x 2 x pi/16 x 0 x 4), e (-i x 2 x pi/16 x 2 x 4), e (-i x 2 x pi/16 x pi/4), e (-i x 2 x pi/16 x 5 i 4), e (-i x 2 x pi/16 x 6 i 4), e (-i x 2 i pi/16 x 7); wherein pi is the meaning of pi, ^ is the power exponent, and e is the meaning of base 2.

In this embodiment, referring to fig. 7, the step S223 mainly relates to the process of obtaining the operation result by the butterfly operation, and may specifically include steps S231-S232, which are respectively described as follows.

Step S231, the butterfly operation controller 123 of the second processing unit 12 in fig. 2 searches the transform data by using the butterfly index corresponding to each hierarchical level according to the calculated hierarchical order to obtain a plurality of groups of index data, and searches the phase factor information by using the butterfly index corresponding to each hierarchical level to obtain the phase factors respectively matched with each group of index data; in addition, the butterfly operation controller 123 multiplies each group of index data by the matched phase factor through butterfly operation to obtain product data corresponding to each stage.

Referring to fig. 10, sample data x (n) = [ x (0), x (1), … … x (14), x (15) ] of 16 points is explained as an example. In fig. 10, p1 is a sample sequence of sample data of 16 points, p2 is a plurality of sets of index data obtained by searching the transform data by the corresponding butterfly index (0, 4,1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11, 15) when stage =2, and the plurality of sets of index data can be represented as: x '(0), x' (4), x ', 1), x', 5), x '(2), x', 6), x '(3), x', 7), x ', 8), x', 12), x '(9), x', 13, x '10, x' 14, x '11, x' 15; wherein the data points of the transformed data are represented by x'. Since x ″ (0), x ″ (4) matched phase factors are e ^ i (— i ^ 2 × pi/16 × 0 ^ 4), x ″ (1), x ″ (5) matched phase factors are e ^ i (— i ^ 2^ pi/16 ^ 1 ^ 4), x ″ (2), x ″ (6) matched phase factors are e ^ i (— i ^ 2^ pi/16 ^ 2^ 4), x ^ (3), x ″ (7) matched phase factors are e ^ i (— i ^ 2^ pi/16 ^ 3 ^ 4), x ″ (8), x ″ (12) matched phase factors are e ^ i ^ 2^ pi/16 ^ 4), x ^ i (9) matched phase factors are e ^ i ^ 4), x ^ i ^ 6 ^ x ^ 4 (13 ^ i ^ 6), and the matched phase factors of x '(11) and x' (15) are e (-i ^ 2^ pi/16 ^ 7 ^ 4), so that the product data corresponding to each classification can be obtained by corresponding multiplication.

In step S232, the butterfly operation controller 123 of the second processing unit 12 in fig. 2 performs butterfly operations sequentially from the highest hierarchy in the hierarchical order to obtain product data corresponding to each hierarchy until iteration is performed to the lowest hierarchy in the hierarchical order, and takes the obtained product data as an operation result.

For example, when the number of data points is 4M =2²²The number of input points is 1024=2¹⁰Since the number of stages of computation is 12, product data corresponding to each stage is obtained by 12 times of butterfly operation from the highest stage 12 until the butterfly operation processing of stage 1 is completed, and the product data obtained 12 times is used as an operation result.

For example, when the number of data points is 16=2⁴The number of input points is 4=2 for one frame of sampling data²Since the number of stages of computation is 2, product data corresponding to each stage is obtained by 2 times of butterfly operation from the highest stage 2 until the butterfly operation processing of stage 1 is completed, and the product data obtained by 2 times is used as an operation result.

In one embodiment, in order toTo clearly understand the technical solution in this embodiment, the number of data points is 16=2 here⁴The number of input points is 4=2 for one frame of sampling data²The FFT accelerated computation kernel of (1) is exemplified for technical explanation.

A1) Setting data = x (n) = [ x (0), x (1), … … x (14), x (15) ], hierarchical stage =2, input to the function search _ index _ function (data, stage), which returns data index information data _ index, see in particular fig. 8.

A2) And searching x (n) by using the data _ index, grouping the index results, sending the index results to an FFT (fast Fourier transform) accelerated computing kernel according to the groups to perform fast Fourier transform, and arranging the transform results according to a natural sequence to obtain transform data.

A3) By phase factor calculation rulesPhase_factor= exp(-i×2π/2^m×(0:(2^(m-k) -1)) ×2^k) And calculating the phase factors corresponding to the k-th grading in the cases of m =4 and k = stage.

A4) Set DATA = [0,1,2 … … 13,14,15], k = stage, input to the function button _ index _ function (DATA, stage), which will return the butterfly index button _ index corresponding to the kth level.

A5) And searching the transformed data in the step A2 by using the button _ index to obtain a plurality of groups of index data, and searching the phase factor in the step A3 to obtain a matched phase factor, so that each group of index data and the matched phase factor can be multiplied by butterfly operation to obtain product data corresponding to the kth grade.

A6) Resetting stage =1, repeating step a3, and continuing to calculate the phase factor corresponding to the k-th stage when m =4 and k = stage. It will be appreciated that k is successively reduced during the iterative computation.

A7) Since stage =1, the step a4 is repeated, and the butterfly index button _ index corresponding to the k-th ranking is calculated continuously if k = stage.

A8) Step a5 is executed again, and calculation of product data corresponding to the k-th hierarchy is continued with k = stage.

A9) Since the butterfly budget has been completed in both cases of the hierarchical order k =2,1, the operation result is obtained by integrating the product data.

In one embodiment, in order to more clearly understand the technical solution in this embodiment, the number of data points is 4M =2 here²²The number of input points is 1024=2¹⁰The FFT accelerated computation kernel of (1) is exemplified for technical explanation.

A1) Setting data = x (n) = [ x (0), x (1), … … x (n-1), x (n)]，n∈[0,2²²-1]The stage =12, and is input to a function search _ index _ function (data, stage), which returns data index information data _ index.

A3) By phase factor calculation rulesPhase_factor= exp(-i×2π/2^m×(0:(2^(m-k) -1)) ×2^k) And calculating the phase factors corresponding to the k-th grading in the cases of m =22 and k = stage.

A4) Set DATA = [0,1,2 … … 2 … … 2²²-3, 2²²-2, 2²²-1]K = stage, and is input to the function button _ index _ function (DATA, stage), which returns the butterfly index button _ index corresponding to the k-th ranking.

A6) Resetting stage = stage-1, repeating step a3, and continuing to calculate the phase factor corresponding to the kth stage when m =22 and k = stage. It will be appreciated that k is successively reduced during the iterative computation.

A7) Since stage = stage-1, the step a4 is repeated, and the butterfly index button _ index corresponding to the k-th ranking is calculated continuously if k = stage.

A9) Repeatedly executing the steps A6-A8 until the butterfly operation is completed when the stage = 1; since the butterfly budget has already been completed in the case of the hierarchical order k =12,11, … 2,1, the operation result is obtained by integrating the product data.

In the present embodiment, referring to fig. 11, in order to facilitate writing of the transform data of the sample data and the butterfly index information into the memory 13, the memory 13 may be partitioned into a first storage space Q1 and a second storage space Q2.

The first storage space Q1 includes a plurality of designated regions corresponding to hierarchical order, such as 0 to (m-k) designated regions, each number representing transformed data at a hierarchy, denoted by a (r), r ∈ [0, (m-k) ]. Each portion a (r) is divisible into 2^ (m-k) subsets, denoted by c (z), z ^ 0, 2^ (m-k) -1; if the number of data points of the sample data is 4M, FFT and the calculation acceleration kernel is 1024, m =22 and k =10 are satisfied here.

The second storage space Q2 includes a plurality of index regions corresponding to hierarchical order, such as 0 to (m-k) -1 index regions, each number representing a butterfly index under a hierarchy, denoted by b (t), t ∈ [0, (m-k) -1 ].

Then, referring to fig. 12, the number of data points will be 4M =2²²The number of input points is 1024=2¹⁰The FFT accelerated computation kernel of (1) is an example for technical description of data storage. Since the butterfly operation needs to be performed through 12 stages, the series relationship between a (r) and b (t) in the case of 12 stages in total is shown in fig. 12. The FFT accelerated computation core 122 in fig. 3 is enabled only when the first processing unit 11 sends data to the second processing unit 12, and the butterfly operation controller 123 reads the butterfly index in b (11) from the 12 th stage to search for a (12), and writes the product data corresponding to the 12 th stage into a (12) after butterfly operation; reading the butterfly index in b (10) to search a (11), writing the product data corresponding to the 11 th stage into a (11) after butterfly operation; and so on, finally reading the butterfly index in b (0) to search a (1), and writing the product data corresponding to the 1 st level into the butterfly operationa (1). It can be understood that the stored data in the first storage space Q1 in the final memory 13 is the operation result for frequency domain analysis.

The technical scheme adopts an FFT (fast Fourier transform) accelerated computing inner core to perform serial computing on the sampled data, so that the situation that the prior parallel processing needs huge multiplication and addition resources can be avoided, and the computing performance of the equipment is improved. In addition, because the butterfly index information and the phase factor information of the sampling data are generated, the data are easily traversed and transformed in an index mode, thereby realizing butterfly operation with the phase factor and further improving the speed of performing time domain-frequency domain transformation on the sampling data.

Example III,

On the basis of the serial FFT co-processing method disclosed in the second embodiment, a digital oscilloscope is disclosed in the present embodiment.

Referring to fig. 13, the digital oscilloscope 3 includes a memory 31 and a processor 32. The memory 31 may be considered as a computer-readable storage medium for storing a program, which may be a program code corresponding to the serial FFT co-processing method in the second embodiment.

The processor 32 is connected to the memory 31 for implementing the serial FFT co-processing method by executing the program stored in the memory 31. The functions performed by the processor 32 may then refer to steps S210-S230 in fig. 4, and to the steps disclosed in fig. 5-7, which will not be described in detail here.

Those skilled in the art will appreciate that all or part of the functions of the various methods in the above embodiments may be implemented by hardware, or may be implemented by computer programs. When all or part of the functions of the above embodiments are implemented by a computer program, the program may be stored in a computer-readable storage medium, and the storage medium may include: a read only memory, a random access memory, a magnetic disk, an optical disk, a hard disk, etc., and the program is executed by a computer to realize the above functions. For example, the program may be stored in a memory of the device, and when the program in the memory is executed by the processor, all or part of the functions described above may be implemented. In addition, when all or part of the functions in the above embodiments are implemented by a computer program, the program may be stored in a storage medium such as a server, another computer, a magnetic disk, an optical disk, a flash disk, or a removable hard disk, and may be downloaded or copied to a memory of a local device, or may be version-updated in a system of the local device, and when the program in the memory is executed by a processor, all or part of the functions in the above embodiments may be implemented.

The present invention has been described in terms of specific examples, which are provided to aid understanding of the invention and are not intended to be limiting. For a person skilled in the art to which the invention pertains, several simple deductions, modifications or substitutions may be made according to the idea of the invention.

Claims

1. A serial FFT coprocessing method is characterized by comprising the following steps:

carrying out conversion processing on the sampled data of the signal by using an FFT (fast Fourier transform) accelerated computing kernel to obtain conversion data;

butterfly index information and phase factor information of the sampling data are generated;

searching the transformation data by using the butterfly index information, and performing butterfly operation on an index result and the phase factor information to obtain an operation result; the operation result is used for frequency domain analysis of the signal.

2. The method of claim 1, wherein the transforming the sampled data of the signal using the FFT-accelerated computation kernel to obtain transformed data comprises:

determining a calculation grading order according to the number of data points of the sampling data and the number of input points of the FFT acceleration calculation kernel;

reordering each data point in the sampled data according to an input data hierarchical ordering rule of the FFT acceleration calculation kernel to obtain data index information corresponding to the highest hierarchy in the calculated hierarchical order;

and searching the sampling data by using the data index information, grouping index results, performing fast Fourier transform according to the groups, and arranging the transform results according to a natural sequence to obtain transform data.

3. The method of claim 2, wherein the generating butterfly index information and phase factor information for the sampled data comprises:

reordering the sampling sequence of the sampling data according to a preset butterfly operation index rule to obtain a butterfly index corresponding to each grading, and integrating butterfly index information of the sampling data;

and taking the sequence of each grading as an independent variable, generating a phase factor corresponding to each grading according to a preset phase factor calculation rule, and integrating to form phase factor information of the sampling data.

4. The method of claim 3,

the butterfly operation index rule comprises: setting positive integer exponential power as step length, combining every two even number serial numbers or every two odd number serial numbers in a sampling sequence of the sampling data, wherein the difference value of the serial numbers in each combination is the step length, and arranging and distributing the formed combinations according to a natural sequence;

the phase factor calculation rule includes: setting the sequence of each grading as k, inputting the formulaPhase_factor= exp(-i×2π/2^m×(0:(2^(m-k) -1)) ×2^k) So as to calculate and obtain a corresponding phase factor; wherein the content of the first and second substances,exp() In order to operate in an exponential manner,iis an imaginary unit, (0 (2)^(m-k)-1)) from 0 to 2^(m-k)-an increment vector of 1, m being the power of the number of data points of said sample data at the base 2.

5. The method as claimed in claim 3, wherein said searching said transform data using said butterfly index information and performing a butterfly operation on an index result and said phase factor information to obtain an operation result comprises:

according to the calculated grading order, searching the transformation data by using the butterfly index corresponding to each grading to obtain a plurality of groups of index data, and searching the phase factor information by using the butterfly index corresponding to each grading to obtain phase factors respectively matched with each group of index data; multiplying each group of index data by the matched phase factor through butterfly operation to obtain product data corresponding to each grading;

and sequentially carrying out butterfly operation from the highest grade in the grading order to obtain product data corresponding to each grade until iteration is carried out to the lowest grade in the grading order, and taking the obtained product data as an operation result.

6. A digital oscilloscope is characterized by comprising a first processing component, a second processing component and a memory which are mutually connected in communication;

the first processing component is used for acquiring sampling data of a signal, performing conversion processing on the sampling data by utilizing an FFT (fast Fourier transform) accelerated computing kernel to generate butterfly index information of the sampling data, and storing the conversion data obtained by the conversion processing and the butterfly index information of the sampling data into the memory;

the second processing component is used for generating phase factor information of the sampling data, searching the transformation data by using the butterfly index information in the memory, performing butterfly operation on an index result and the phase factor information, processing to obtain an operation result, and storing the operation result into the memory;

the first processing unit is further configured to obtain the operation result from the memory, and perform frequency domain analysis on the operation result to form an analysis result.

7. The digital oscilloscope of claim 6, wherein the first processing component comprises a computation grading module, a first data sorting module, a second data sorting module, and a transform processing module;

the calculation grading module is used for determining a calculation grading order according to the number of data points of the sampling data and the number of input points of the FFT acceleration calculation kernel;

the first data sorting module is used for re-sorting each data point in the sampled data according to the input data hierarchical sorting rule of the FFT accelerated computing kernel to obtain data index information corresponding to the highest hierarchical level in the computed hierarchical order;

the transformation processing module is used for searching the sampling data by using the data index information, grouping index results and performing fast Fourier transformation according to groups to obtain a transformation result corresponding to each group;

and the second data sorting module is used for re-sorting the sampling sequence of the sampling data according to a preset butterfly operation index rule to obtain a butterfly index corresponding to each grade.

8. The digital oscilloscope of claim 7, wherein a first storage space and a second storage space are provided in the memory;

the first storage space comprises a plurality of designated areas corresponding to the hierarchical order, the designated areas are used for sequentially storing the transformation results corresponding to the groups generated by the transformation processing module, and the transformation data are formed by arranging the natural order of the data stored in the designated areas;

the second storage space comprises a plurality of index regions corresponding to the hierarchical order; the index areas are used for sequentially storing each butterfly index corresponding to the grades generated by the second data sorting module, and the butterfly index information is formed by the natural sequence arrangement of the written data in each index area.

9. The digital oscilloscope of claim 7, further comprising a display; the display is connected with the first processing component and used for displaying the analysis result formed by the first processing component.

10. A computer-readable storage medium, characterized by comprising a program executable by a processor to implement the method of any one of claims 1-5.